I NTRODUCTION Patients in RCTs may switch treatments for reasons - - PowerPoint PPT Presentation

▶
i ntroduction
SMART_READER_LITE
LIVE PREVIEW

I NTRODUCTION Patients in RCTs may switch treatments for reasons - - PowerPoint PPT Presentation

Nov 14, 2023 280 likes •453 views

Evaluation of methods that adjust for treatment switching in clinical trials Richard Fox a , Lucinda Billingham a b , Keith Abrams c a Cancer Research UK Clinical Trials Unit, University of Birmingham, UK b MRC Midland Hub for Trials Methodology

slide-1
SLIDE 1

Evaluation of methods that adjust for treatment switching in clinical trials

Richard Fox a, Lucinda Billingham a b, Keith Abrams c

a Cancer Research UK Clinical Trials Unit, University of Birmingham, UK b MRC Midland Hub for Trials Methodology Research, University of Birmingham, UK c Centre for Biostatistics and Genetic Epidemiology, University of Leicester, UK

slide-2
SLIDE 2

2

INTRODUCTION

 Patients in RCTs may switch treatments for reasons

associated with their illness

 Treatment switching dilutes estimates of treatment efficacy  From a review of trials featuring treatment switching we

  • bserved 84% switching in one trial

 We investigate this scenario (switch control to intervention)

Randomised Control Switch to intervention Progression Intervention

slide-3
SLIDE 3

3

METHODS

 9 methods identified  Hazard (6) and Time Ratio (3) scales

 Time ratios

 Measure of treatment effect  Extent survival time is modified by treatment  e.g. TR=2 implies survival time doubled on average

 Some methods have numerous test / assumptions

slide-4
SLIDE 4

4

METHODS - HAZARD RATIO SCALE

1.

Intention to treat (ITT)

2.

Per-protocol

i.

Delete patients that switch (PPD)

ii.

  • r Censor at time of switch (PPC)

3.

Time varying covariate (TVC)

4.

Adjusted Cox Model (AdjCox)

Law & Kaldor 1996 1 5.

Causal PH estimator (CaPH)

Loeys & Goetghebeur 2003 2 6.

Inverse Probability Treatment Weighting (IPTW)

Hernan et al 2000 3

slide-5
SLIDE 5

5

METHODS - TIME RATIO SCALE

1.

Rank preserving structural failure time model (RPSFT)

i.

Multiple tests (*4)

Robins & Tsiatis 1991 4 2.

Iterative parameter estimation (IPE)

Branson & Whitehead 2002 5 3.

Parametric randomisation based method (PRB)

Walker et al 2004 6

slide-6
SLIDE 6

6

SIMULATING SURVIVAL DATA

 Survival data has Weibull distribution (λ=1.01, γ=0.5)

 AFT property: dividing log HR by shape parameter (γ) returns

acceleration parameter (TR) 7 e.g. exp(-ln(HR)/γ) = exp(-ln(0.7)/0.5) = 2.04

 Patients randomised to control or intervention (1:1)

 Controlled parameters creating 24 scenarios

 Treatment effect  % good vs bad prognosis within arm  P(switching | prognosis)  Survival / Switching times scaled by prognosis  Censoring %

 Review of NICE technology appraisals informed the above

slide-7
SLIDE 7

7

SWITCHING TRIGGER

 IPTW method requires a time dependent covariate related to

treatment switching / compliance

 Designed biomarker level

Δ ≥ 20%triggers a switch (from baseline)

Beyond switching time level fluctuates around Δ = 20%level

Non-switching patients Δ < 20%

0% 5% 10% 15% 20% 25% % Change Months Switching patient Compliant patient

slide-8
SLIDE 8

8

ASSESSING METHODS

 % Bias  Standard-error of the effect-size  Mean Square Error (combines above)  Coverage

% estimates where 95% CI includes true effect size

 % successful estimates  Averaged over 1000 simulated datasets for each scenario

100 * ) ˆ (    

i

฀  ( ˆ 

i  )2  SE( ˆ



i)2

฀  SE( ˆ 

i)

slide-9
SLIDE 9

9

RESULTS – LOW BIAS SCENARIOS

True effect (HR 0.7)

  • 24%
  • 20%

11% 190% 201% 66% 14%

  • 33%
  • 6%

4% 41% 38% 11% 6% ITT PP - Delete PP - Censor TVC ADJCox CaPH IPTW Scenario 1 (Low Switching) Scenario 2 (Low Switching) True effect (TR 2.04)

  • 11%
  • 4%
  • 6%
  • 3%
  • 3%
  • 3%
  • 5%
  • 3%
  • 10%
  • 7%
  • 7%
  • 6%

RPSFT - Wei RPSFT - LR RPSFT - Cox RPSFT - Exp IPE PRB Scenario 1 (Low Switching) Scenario 2 (Low Switching)

Scenario (Switching) % Good prognosis within treatment group P(switch|prognosis) Effect size (HR / TR) Poor prognosis Good prognosis 1 (Low) 75% 0.5 0.25 0.7 /2.04 2 (Low) 50% 0.5 0.25 0.7 /2.04

HAZARD RATIO SCALE TIME RATIO SCALE

slide-10
SLIDE 10

10

RESULTS – HIGH BIAS SCENARIOS

True effect (HR 0.5)

  • 12%
  • 17%

8% 58% 48% 14% 15%

  • 6%
  • 50%

20% 217% 201% 71% 28% ITT PP - Delete PP - Censor TVC ADJCox CaPH IPTW Scenario 3 (High Switching) Scenario 4 (High Switching) True effect (TR 4)

  • 4%
  • 7%
  • 12%
  • 8%
  • 8%
  • 7%
  • 7%
  • 3%
  • 8%
  • 6%
  • 6%
  • 5%

RPSFT - Wei RPSFT - LR RPSFT - Cox RPSFT - Exp IPE PRB Scenario 3 (High Switching) Scenario 4 (High Switching)

Scenario (Switching) % Good prognosis within treatment group P(switch|prognosis) Effect size (HR / TR) Poor prognosis Good prognosis 3 (High) 50% 0.85 0.25 0.5 / 4 4 (High) 75% 0.85 0.5 0.5 / 4

HAZARD RATIO SCALE TIME RATIO SCALE

slide-11
SLIDE 11

11

RESULTS – ERRATIC RESULTS

 PRB can return erratic results

 Sensitive to specification of frailty

Scenario % Good prognosis within treatment group P(switch|prognosis) Effect size (HR / TR) Poor prognosis Good prognosis Morden8 - Sc 6 30% 0.75 0.5 0.7 / 2.04

True effect (TR 2.04)

  • 2%
  • 3%
  • 7%
  • 3%
  • 3%
  • 3%

RPSFT - Wei RPSFT - LR RPSFT - Cox RPSFT - Exp IPE PRB Erratic PRB method

slide-12
SLIDE 12

12

RESULTS – KEY FINDINGS

Switching is common in clinical trials

ITT results can be heavily biased

Per-protocol is not appropriate where switching occurs

Adjustment not routinely applied

Some of the methods available (Stata)

RPSFT and IPE consistent under these conditions

IPE has 100% successful estimation

IPE also returns estimates of the Weibull parameters

Results robust to additional censoring

slide-13
SLIDE 13

13

CONCLUSION

We recommend that the IPE method of Branson & Whitehead be utilised in the analysis of clinical trials that feature treatment switching. Available from Ian White’s software page: http://www.mrc-bsu.cam.ac.uk/Software/stata.html#Software_IW My email: foxrp@bham.ac.uk

slide-14
SLIDE 14

14

EXTENSIONS

 Simulate alternative survival distributions  Additional covariates  Multiple switching directions  Dependent censoring  Other methods

 Meta / Bayes analysis  Structural nested mean models

 Statistical analysis plan – sensitivity

slide-15
SLIDE 15

15

REFERENCES

1.

Law and J Kaldor. Survival analyses of randomised clinical trials adjusted for patients who switch

  • treatment. Stat Med, 15:2069-2076, 1996.

2.

T Loeys and E Goetghebeur. A causal proportional hazards estimator for the effect of treatment actually received in a randomised trial with all-or-nothing compliance. Biometrics, 59(1):100-105, 2003.

3.

M Hernan, B Brumback, and J Robins. Marginal structural models to estimate the causal effect of zidovudine on the survival of hiv-positive men. Epidemiology, 11(5):561-570, September 2000.

4.

J Robins and A Tsiatis. Correcting for non-compliance in randomised trials using rank preserving structural failure time models. Communication in Statistics-Theory and Methods, 20(8):2609-2631, 1991.

5.

M Branson and J Whitehead. Estimating a treatment effect in survival studies in which patients switch treatment. Stat Med, 21:2449-2463, 2002.

6.

S Walker, I White, and A Babiker. Parametric randomization-based methods for correcting for treatment changes in the assessment of the causal effect of treatment. Stat Med, 23:571-590, 2004.

7.

Collett, D. (2003), Modelling Survival Data in Medical Research (2nd ed.)

8.

J Morden et al. Assessing methods for dealing with treatment switching in randomised controlled trials: a simulation study. BMC Med Res Methodol. 2011 Jan 11;11:4.